March 18, 2014
The PFM, My Bid Limits, and You
Throughout the positional tier series and continuing throughout my bid limit articles, some of our readers have been asking questions and raising objections along the following lines:
I don’t understand why your bid limit on Paul Goldschmidt is $36 in NL-only when the PFM only has him at $29.20. I know everyone is really excited about Goldschmidt this year, but I’m relying on the PFM for my auction and I don’t understand why there is a $7 difference between your suggested bid and the PFM’s. What gives?
For more than a few readers, I suspect there is a misunderstanding of what PECOTA/the PFM is designed to do versus what my bid limits are designed to do.
(Before you dive into this piece, I strongly urge you to read Rob McQuown’s 2014 introduction to PECOTA if you have general questions about PECOTA and the PFM. My article is not intended to reinvent the wheel when it comes to explaining PECOTA and the PFM, but rather to showcase the difference between PECOTA and my bid limits.)
As noted in Baseball Prospectus’s glossary, PECOTA has three elements incorporated into its projection model:
1) Major-league equivalencies, to allow us to use minor-league stats to project how a player will perform in the majors;
The most important component to consider for fantasy valuation is item #2. The forecasts offer projections that incorporate regression to the mean and weighted averages for every player. For a projection system this is appropriate; in fact, it would be a poor projection system if it predicted something silly like “Chris Davis is going to hit 53 home runs” or “Miguel Cabrera is going to have a .348 batting average.” Yet this is exactly what happened in 2013. Something PECOTA will not do is project the inevitable outliers and – to reiterate this vital point – it most definitely should not.
Nearly every fantasy analyst is not factoring regression to the mean and weighted averages into his bid limits, and my bid prices are no exception. Rather, we are all attempting to offer recommendations based upon the market price for the best hitters and pitchers available at your auction.
Table 1: Top 10 AL-Only and NL-Only Hitters: PFM versus LABR
According to PECOTA, Braun and Trout are projected to have the best fantasy seasons in the National League and American League. But Braun is “only” projected to earn $35 while Trout is “only” projected to earn $37. A $37 projection is certainly a reasonable one for Trout given that it’s only six dollars less than what the PFM said he earned last year. But LABR doesn’t know from PECOTA. LABR’s $45 bet is saying it believes that a 22-year-old stud hitter will continue to get better as he moves up the age curve.
This is only part of what LABR’s price reflects. LABR recognizes that Trout could put up a $37 season, as the PFM suggests. To take this even a step further, LABR recognizes that Trout could have a down year and even if he stayed completely healthy earn only $25-30. Trout’s buyer understands the fact that there is a good chance that Trout won’t have a $45 season. The bet here isn’t on Trout having a $45 season but rather on the idea that Trout is the most likely hitter in the American League to return $35 or more worth of statistics.
While this principle isn’t quite as extreme with the other most expensive hitters, it still applies. Nearly every hitter in this bracket is being paid for his ceiling, and PECOTA is correct that there is a good chance that nearly all of these hitters will fail to get there.
Table 2: Top 10 Salaries, AL/NL Only Hitters, 2009-2012
Table 2 shows the 10 most expensive hitters using the average salary from three expert leagues—CBS, LABR, and Tout Wars—from 2009 through 2013. The expert market pays these hitters with the expectations that will perform as well or slightly better than they did the year before, but the harsh reality is that these hitters disappoint their fantasy owners year in and year out, losing about $10 per hitter on average for the last five years.
PECOTA might miss on individual hitters on a case-by-case basis, but it has a far better handle on the big picture than the fantasy experts do. If this is the case, why should you bother with my bid limits? Doesn’t this exercise simply prove that my bid limits are worthless and that the PFM is more accurate?
Remember that part of the operating premise of this article is that as a prognostication tool PECOTA works quite well and is not what is at issue here. PECOTA looks at the future and correctly predicts regression for the top hitters on a macro level. Next, the PFM takes PECOTA’s forecast and applies it to fantasy values, also doing a good job based on the stats that PECOTA generates. I quibble with how much value the PFM gives to stolen bases and saves (even with the SGP option turned on) and believe that the mono-league formulas give too much credence to the concept of the replacement level fantasy player, but the dollar values are generally sound. A $2 difference between my 2013 Paul Goldschmidt valuation and the PFM’s 2013 Goldschmidt isn’t a gaping chasm.
My bid limits are different than the PFM valuations not because PECOTA is right and I’m wrong, but because we are serving two different masters. Where PECOTA is trying to construct a reality based upon a regression-based data model, I have devised a system that is attempting to navigate the dynamics of a fantasy baseball auction. My bids do something that the PFM cannot: adjust for the real world dynamics of a fantasy baseball auction.
Since 2011, Miguel Cabrera has earned $36, $40, and $42 according to my valuations (the PFM says $31, $39, and $41). I look at these earnings, see a 31-year-old four-category deity, and put my bid limit at $42. Am I betting on a repeat of 2013? No. In fact, if I ran my own mechanical projections I would probably show Cabrera putting up a $35-37 season. So why am I willing to possibly take a $5-7 loss on Cabrera?
I look back at retrospective fantasy baseball earnings and know between seven to fifteen hitters will earn $30 or more in mono formats. My bids are guessing which hitters are most likely to earn $30 or more. I know that I’m going to get some of my guesses wrong, but because anywhere between 7-15 hitters are going to earn $30 or more, my bids are structured so that about 10-12 hitters have a bid limit of $30 or higher.
This is what the auction marketplace does as well. The expert market doesn’t look at PECOTA or any of the non-Baseball Prospectus projection systems and shy away from spending over $30 on all but Trout and Cabrera. It places bets on the hitters it believes are most likely to crack the $30 barrier, and bets based on what happened last year, not what might happen this year. The market is aware that it is going to miss on some hitters (and miss big on a few hitters) but doesn’t care because the market isn’t trying to factor in regression but is trying to buy statistics.
Despite the big losses among the most expensive hitters year in and year out, they still earn the most money nearly every season. You need value and profit to win but you have to procure enough statistics. This is why the market overpays for the top hitters. Because the market is overpaying the best hitters, the hitters at the end of the auction are relative bargains. Some one-dollar hitters are going to flame out and lose money, but all it took in 2013 was Brian Dozier to skew the profit margins for the cheap hitters way up.
If you religiously use the PFM’s values in an auction, it is likely you will struggle to purchase players early, get a lot of players in the middle of your auction, and end the day leaving some money unspent. The PFM is “paying” the full freight for what certain players might do if they play full time, while the auction market is conservative. For example, perhaps the PFM is correct and Josh Rutledge will earn $16.72 in 2014. It’s certainly possible that he can do this if he wins the second base job over D.J. LeMahieu, but his purchase prices thus far have been three dollars (CBS) and seven dollars (LABR). If want Rutledge badly enough and believe my five dollar bid limit is ridiculously conservative, you should push it higher but $17 is way too much. No one is going to pay that much for Rutledge. $10-12 should do it.
I have been doing this for so long that I don’t need to look at a projection system before I put my bids together. I simply do what I did above with Miguel Cabrera above for most of the available player pool. I review recent earnings, guess what I believe the player will earn this year based on a combination of fantasy earnings, sabermetric data, and any changes in his real-life circumstances and make an educated guess. Then in cases like Rutledge’s I move the bid down to reflect what I think the player should actually cost in an auction-style marketplace.
Hopefully, I have convinced you not to go into your auctions solely using the PFM valuations and to use my bid limits as a baseline instead. However, if you still believe that my bid limits contain too much of the human element and adamantly believe that the PFM is superior, I recommend making the following adjustments.
1) Turn the SGP feature ON.
If you don’t do this, one-dimensional stolen base threats and closers will be valued extremely highly and there’s a good chance that you will walk out of your auction with Billy Hamilton, Eric Young, Craig Kimbrel, and Kenley Jansen at a combined salary of $100 or more. You’ll win saves and steals going away but will fail to acquire the balance that you need to win.
2) Don’t Pay The Bench Guys
This applies even more for middle relievers. Some middle reliever is surely going to do what Neal Cotts did last year and earn $15 or more in middle relief, but there’s no reason to pay more than $1-2 for a middle reliever when there are so many relievers sitting in the free agent pool, and trying to predict who the best middle relievers will be in fantasy is an exercise in pure guesswork.
I would recommend taking the money from the bottom and moving it to the top, but if you agree with PECOTA and are afraid of regression, move that money to the middle. Just make sure you’re not wasting your bid money on bench players you can get for $2-3 at the end quite easily.
3) Make Sure You Spend All Of Your Money
You don’t want to overspend on one of top hitters, but it’s even worse to panic and pay $40 for Puig because he was the last top-tier player on the board (in the PFM’s model) and you didn’t want to get caught with a lot of money at the end of the auction. My bid limits ensure that you will spend your money in your auction. Even if you don’t agree with my bid limits for specific players, take the structure of the bid limits (the pricing hierarchy) and use them as a jumping off point. Points #2 and #3 are closely tied together. Even in an only league, there are going to be anywhere from 12-24 one dollar hitters and 10-15 one dollar pitchers in the end game. Make sure that your bid limits reflect how teams and owners spend in your league’s auction dynamic.
You are not competing against PECOTA or the PFM but against the other owners in your league. The bid limits I have constructed have the built-in assumption that most Rotisserie-style leagues are similar to CBS, LABR, and Tout Wars in the way that they allocate their auction dollars. Feel free to disagree with my rankings on a player-by-player basis, but the logic behind how I distribute the bids across the player pool is sensible and borne out by my real life Roto auction experience.